Fortifying the barrier: the impact of lipid A remodelling on bacterial pathogenesis.

Gram-negative bacteria decorate their outermost surface structure, lipopolysaccharide, with elaborate chemical moieties, which effectively disguises them from immune surveillance and protects them from the onslaught of host defences. Many of these changes occur on the lipid A moiety of lipopolysaccharide, a component that is crucial for host recognition of Gram-negative infection. In this Review, we describe the regulatory mechanisms controlling lipid A modification and discuss the impact of modifications on pathogenesis, bacterial physiology and bacterial interactions with the host immune system.

The bacterial cell envelope is a complex structure that protects the cell from the surrounding environment. A defining feature of Gram-negative bacteria is the presence of an outer membrane, which is an asymmetrical bilayer with glycerophospholipids confined to the inner leaflet and lipopolysaccharide (LPS) anchored to the outer leaflet[1] (FIG. 1a,b). Similarly to most cell envelope components, LPS is made at the cytoplasmic face of the inner membrane and must be transported across the two bilayers and the periplasm to become integrated in the outer membrane[1].

Figure 1 |

The cell envelope of Gram-negative bacteria.

a | A cryoelectron tomography image of an Escherichia coli cell, showing the characteristic inner membrane (IM) and outer membrane (OM) (scale bar of 200 nm)[123]. b | Schematic of the Gram-negative cell envelope, showing the typical inner and outer bilayers that are separated by the periplasm, which contains peptidoglycan (PG). The outer leaflet of the outer membrane contains lipopolysaccharide (LPS), which is anchored to the membrane by the LPS lipid A domain[1]. The inner leaflet of the outer membrane and also the entire inner membrane are composed of phospholipids only, and both bilayers can contain a range of different types of membrane protein. c | The lipid A and inner core (Kdo (3-deoxy-d-manno-octulosonic acid)) portion of LPS are shown. Unmodified lipid A consists of a β-1′,6-linked disaccharide of glucosamine that is both phosphorylated and fatty acylated[7]. This basic structure can be extensively modified after synthesis. d | The LPS modifications that occur in Salmonella spp. The enzymes responsible are controlled by either the PmrAB two-component system (red) or the PhoPQ two-component system (blue), or have no known two-component regulatory system (green). The various possible modifications include the addition of 4-amino-4-deoxy-l-arabinose (aminoarabinose) moieties (by ArnT) and phosphoethanolamine moieties (by EptA and EptB), as well as phosphorylation (by LpxT), deacylation (by PagL and LpxR[7]; resulting in loss of acyl chains, as indicated by dashed lines), acylation (by PagP)[7] and hydroxylation (by LpxO). Transcription of the gene encoding LpxO is modestly induced by PhoPQ, but LpxO remains active in conditions in which PhoPQ is inactive, suggesting that the enzyme acts independently of this two-component system[124].

LPS is composed of three domains: the lipid A hydrophobic anchor, the core oligosaccharide and the O antigen[1] (FIG. 1b). Some organisms (for example, Neisseria spp.) produce lipooligosaccharide (LOS), in which the repeating O antigen domain is absent and is replaced by an extended core region[2]. Lipid A, the endotoxic portion of LPS and the site for many LPS modifications, is initially synthesized as a β-1′,6-linked disaccharide of glucosamine that is both phosphorylated and fatty acylated (FIG. 1c). In some organisms, such as Escherichia coli K12, this structure represents the typical form of lipid A in the outer membrane. Despite initial studies reporting that lipid A could be modified with polar substituents (such as amino sugars[3]), it was nevertheless viewed as a static structure. This view changed in the late 1990s following the characterization of the lipid A biosynthesis pathway, which established a foundation for studies which discovered that lipid A is altered after synthesis[4-6]. In fact, Gram-negative organisms have evolved several LPS modification strategies that allow these organisms to adapt to their unpredictable and often hostile surroundings[7]. During and after trafficking to the cell surface, lipid A can undergo extensive remodelling, resulting in the wide variety of lipid A structures that are observed across species (FIG. 1d). This is accomplished by diverse lipid A modification enzymes that remove or add acyl chains and phosphate groups, as well as other enzymes that transfer various constituents onto the molecule, such as sugars, phosphoryl containing groups and even an amino acid, all of which can influence bacterial interactions with the host. The term lipid A modification is used throughout this Review to refer to biological modification of lipid A in the bacterial cell as well as artificial modification in vitro. Modifications also occur on other moieties of LPS (BOX 1), but these non-lipid A modifications are not the focus of this Review.

Lipid A alterations directly affect pathogenesis by changing outer-membrane permeability, promoting resistance to antimicrobial peptides and interfering with the ability of the host to recognize LPS as a conserved microorganism-associated molecular pattern (MAMP)[7]. The diversity of LPS modification systems is quite extraordinary, and the accumulated knowledge about these systems has provided deeper insight into bacterial mechanisms that contribute to human disease and immunity. In this Review, we describe the mechanisms that regulate lipid A remodelling, the host defence systems that recognize this molecule and the strategies that bacteria have evolved to use lipid A modifications for their own benefit. Finally, we discuss the interplay between lipid A modification systems and bacterial physiology.

Regulation of lipid A remodelling

Lipid A modifications are often necessary only for a portion of the bacterial life cycle, such as host colonization, and as a consequence the enzymes responsible for the modifications are subjected to both transcriptional and post-translational regulation. Many modification enzymes are embedded in the outer membrane in close proximity to lipid A, necessitating tight control for selective enzyme activity, whereas other enzymes are constitutively active regardless of their localization. Two-component systems, small RNAs (sRNAs), peptide feedback loops and substrate availability are all involved in directing the activity of these enzymes (FIG. 2).

Figure 2 |

Transcriptional and post-translational regulation of lipid A modification enzymes.

a | Transcriptional control of lipid A modification enzymes includes gene regulation by two-component systems such as PhoPQ, leading to acylation and deacylation of lipid A by upregulating transcription of the genes encoding the enzymes PagP and PagL, respectively. The two-component system PmrAB leads to the addition of 4-amino-4-deoxy-l-arabinose (aminoarabinose; l-Ara4N) and phosphoethanolamine (pEtN) to lipid A by upregulating transcription of the genes encoding the innermembrane enzymes ArnT and EptA, respectively, which modify lipid A as it is transported to the outer membrane[10,11]. Expression of EptB (another phosphoethanolamine transferase) is repressed by the small RNA (sRNA) MgrR, which is induced by PhoPQ[30]. Translation of the phoP mRNA is repressed by the sRNA MicA[31], which leads to the loss of regulation by the PhoPQ system. The sRNA MicF increases degradation of the lpxR mRNA, which encodes a lipid A deacylase. b | Post-translational control of lipid A modification systems includes inhibition of the kinase LpxT (which phosphorylates lipid A during transport to the outer membrane) by the small peptide PmrR, which is upregulated by the PmrAB system in response to high levels of Fe3+ (REF. 33). Post-translational regulation is also mediated by substrate availability. Membrane perturbation can lead to the displacement of phospholipids from the inner leaflet to the outer leaflet of the outer membrane, placing these donor substrates in close proximity to the acyltransferase PagP, and thus enhancing enzyme activity. PagP cleaves the phospholipid substrate, restoring the composition of the outer membrane and increasing the integrity of the permeability barrier by further acylating lipid A[39]. LpxR deacylates lipid A, but this activity is inhibited by the aminoarabinose lipid A modification.

Transcriptional control

Two-component systems are typically composed of a sensor kinase and a response regulator; the sensor kinase is autophosphorylated on stimulation and transfers its phosphate group to the response regulator, which then serves as a transcription factor. To date, various two-component systems have been implicated in the regulation of lipid A modification enzymes, including the widespread PhoPQ and PmrAB (also known as BasRS) systems, and the ParRS and CprRS systems in Pseudomonas aeruginosa[8-13]. Other systems, such as EvgAS, RcsBC and BvgAS, have been linked to the activity of the PhoPQ system or to downstream PhoPQ-regulated genes, but their direct involvement in the modification of lipid A has not yet been thoroughly investigated[14-17]. The PhoPQ and PmrAB systems have been reviewed extensively elsewhere[10,11], so a brief overview is provided here with a focus on the more recent discoveries.

Functional PhoPQ systems are widespread among bacteria, although there is some interspecies variation in the activation signal used and the genes that are regulated[10,18]. Activation of the PhoPQ system in Salmonella enterica subsp. enterica serovar Typhimurium by acidic pH, certain antimicrobial peptides, and the depletion of Mg2+ and Ca2+ stimulates transcription of pagP and pagL and subsequent upregulation of the encoded proteins, which acylate and deacylate lipid A, respectively[9,10,19-21] (FIGS 1d,2; TABLE 1). Furthermore, both of these enzymes are subjected to post-translational regulation (see below). Because the active sites of these two enzymes are found on the extracellular surface of the outer membrane and in close proximity to lipid A, they require tight control to ensure that lipid A modification is appropriately regulated.

Table 1|

Modifications to the Kdo–lipid A domain of lipopolysaccharide

Enzyme*	Enzyme localization	Active-site topology	Pathogenic organisms[‡]	Activity	Contributors to regulation	Effect of modification	Refs
AlmG	Inner membrane	Cytoplasmic	V. cholerae	Transfer of glycine to the hydroxyl group of the 3′-acyloxyacyl chain of lipid A	Unknown	CAMP resistance	75
ArnT (PmrK)	Inner membrane	Periplasmic	E. coli, P. mirabilis, S. Typhimurium, S. flexneri, P. aeruginosa, B. cepacia, Y. enterocolitica, Y. pestis, Y. pseudotuberculosis, K. pneumoniae, B. pertussis and B. bronchiseptica	Addition of aminoarabinose to lipid A	PmrAB, as well as ParRS and CprRS in P. aeruginosa	CAMP resistance	7,14,125–128
EptA (LptA and PmrC)	Inner membrane	Periplasmic	E. coli, V. cholerae, H. pylori, S. Typhimurium, S. flexneri, Neisseria gonorrhoeae and Neisseria meningitidis	Addition of phosphoethanolamine to lipid A	PmrAB	CAMP resistance	7,125,129,130
EptB	Inner membrane	Periplasmic	E. coli, S. Typhimurium and Y. pestis	Transfer of phosphoethanolamine to Kdo	PhoPQ and the sRNA MgrR	Modest CAMP resistance	125,131,132
EptC	Inner membrane	Periplasmic	C. jejuni	Transfer of phosphoethanolamine to lipid A, the flagellar rod and other substituents	Present under normal laboratory conditions	CAMP resistance and motility	91,133
FlmF1	Inner membrane	Periplasmic	F. tularensis	Addition of glucose and mannose to lipid A	Present under normal laboratory conditions	Possible role in CAMP resistance	134
FlmF2	Inner membrane	Periplasmic	F. tularensis	Addition of galactosamine to lipid A	Present under normal laboratory conditions	Possible role in CAMP resistance	134,135
FlmK	Inner membrane	Periplasmic	F. tularensis	Addition of glucose, mannose or galactosamine to lipid A	Present under normal laboratory conditions	TLR4 evasion and a possible role in CAMP resistance	135
KdkA	Inner membrane	Cytoplasmic	H. influenzae, V. cholerae, P. multocida, B. pertussis, A. actinomycetemcomitans and S. putrefaciens	Phosphorylation of Kdo	Present under normal laboratory conditions	Possible effect on toxin delivery	125,136
KdoH1 and KdoH2 (KdhA)	Inner membrane	Cytoplasmic	H. pylori, F. tularensis and L. pneumophila	Removal of outer Kdo region	Present under normal laboratory conditions	CAMP resistance	137–139
KdoO	Inner membrane	Cytoplasmic	B. cepacia, B. ambifaria, Y. pestis and A. haemolyticus	Hydroxylation of Kdo	Present under normal laboratory conditions	Unknown	140
LmtA	Inner membrane	Cytoplasmic	L. interrogans	Methylation of lipid A	Present under normal laboratory conditions	Unknown	141
LpxD2	Inner membrane	Cytoplasmic	F. tularensis	Addition of shorter (C16) primary acyl chains to lipid A at low temperatures	Low temperatures	Increased membrane fluidity	72
LpxE	Inner membrane	Periplasmic	F. tularensis, H. pylori and P. gingivalis	Dephosphorylation of lipid A	Present under normal laboratory conditions	CAMP resistance and TLR4 evasion	7,125,142
LpxF	Inner membrane	Periplasmic	F. tularensis, H. pylori, P. gingivalis and L. interrogans	Dephosphorylation of lipid A	Present under normal laboratory conditions	CAMP resistance and TLR4 evasion	7,63,125,143,144
LpxO	Inner membrane	Cytoplasmic	S. Typhimurium, K. pneumoniae, P. aeruginosa, B. bronchiseptica and L. pneumophila	Hydroxylation of lipid A acyl chains	Present under normal laboratory conditions	Coordination of stress responses	7,125,145
LpxP	Inner membrane	Cytoplasmic	E. coli, S. Typhimurium, Y. enterocolitica, Y. pestis, and L. pneumophila	Alternative biosynthetic acyltransferase at low temperatures	Low temperatures	Membrane integrity and fluidity	7,146,147
LpxR	Outer membrane	Extracellular	E. coli O157:H7, H. pylori, V. cholerae, S. Typhimurium, Y. enterocolitica, Y. pestis and Y. pseudotuberculosis	Deacylation of lipid A	The sRNA MicF and aminoarabinose modifications to lipid A	TLR4 evasion	7,125,148
LpxT	Inner membrane	Periplasmic	E. coli, S. Typhimurium and Y. pestis	Phosphorylation of lipid A and recycling of undecaprenyl pyrophosphate	PmrAB and the small peptide PmrR	Unknown	34
PagL	Outer membrane	Extracellular	P. aeruginosa, B. cepacia, S. Typhimurium, B. pertussis and B. bronchiseptica	Deacylation of lipid A	PhoPQ and aminoarabinose modifications to lipid A	Decreased TLR4 activation	7,125,128,149
PagP	Outer membrane	Extracellular	E. coli, S. Typhimurium, S. flexneri, P. aeruginosa, Y. enterocolitica, Y. pestis, Y. pseudotuberculosis, K. pneumoniae, B. pertussis, B. bronchiseptica, B. parapertussis and L. pneumophila	Acylation of lipid A	PhoPQ and membrane perturbation	Selective CAMP resistance, membrane integrity and decreased TLR4 activation	7,125,150–152

A. haemolyticus, Acinetobacter haemolyticus; A. actinomycetemcomitans, Actinobacillus actinomycetemcomitans; aminoarabinose, 4-amino-4-deoxy-l-arabinose; B. bronchiseptica, Bordetella bronchiseptica; B. parapertussis, Bordetella parapertussis; B. pertussis, Bordetella pertussis; B. ambifaria, Burkholderia ambifaria; B. cepacia, Burkholderia cepacia; CAMP, cationic antimicrobial peptide; C. jejuni, Campylobacter jejuni; E. coli, Escherichia coli; F. tularensis, Francisella tularensis; H. influenzae, Haemophilus influenzae; H. pylori, Helicobacter pylori; Kdo, 3-deoxy-d-manno-octulosonic acid; K. pneumoniae, Klebsiella pneumoniae; L. pneumophila, Legionella pneumophila; L. interrogans, Leptospira interrogans; N. gonorrhoeae, Neisseria gonorrhoeae; N. meningitidis, Neisseria meningitidis; P. multocida, Pasteurella multocida; P. gingivalis, Porphyromonas gingivalis; P. mirabilis, Proteus mirabilis; P. aeruginosa, Pseudomonas aeruginosa; S. Typhimurium, Salmonella enterica subsp. enterica serovar Typhimurium; S. putrefaciens, Shewanella putrefaciens; S. flexneri, Shigella flexneri; TLR4, Toll-like receptor 4; V. cholerae, Vibrio cholerae; Y. enterocolitica, Yersinia enterocolitica; Y. pestis, Yersinia pestis; Y. pseudotuberculosis, Yersinia pseudotuberculosis.

Alternative enzyme names in certain species are given in brackets.

Organisms containing homologues of these enzymes are excluded from the table if the respective modification has not been observed. The list of enzymes in this table is likely to expand as the Kdo–lipid A region is further characterized in a range of species.

In S. Typhimurium, PhoPQ further influences lipid A modification by activating the PmrAB system, although in E. coli and P. aeruginosa the two systems are not cou-pled[22]. Direct activation of PmrAB in S. Typhimurium occurs upon sensing Fe3+, Al3+ and low pH, and leads to the upregulation of genes such as arnT (also known as pmrK) and eptA (also known as lptA and pmrC), which transfer 4-amino-4-deoxy-l-arabinose (aminoarabinose) and phosphoethanolamine groups to lipid A, respectively[11] (FIGS. 1d,2a). Similarly to the S. Typhimurium PmrAB system, the P. aeruginosa PmrAB system is activated by the depletion of cations (such as Mg2+), but it is also activated by antimicrobial peptides[23]. In E. coli, S. Typhimurium, P. aeruginosa and other bacteria, the activity of the enzymes controlled by PhoPQ and PmrAB strengthen the integrity of the outer-membrane permeability barrier in the presence of antimicrobial peptides and depleted cations, thus enhancing bacterial survival in the host[24,25].

The ParRS and CprRS are independent two-component systems that have been found only in P. aeruginosa and respond to subinhibitory concentrations of antimicrobial peptides[12,13]. Thus, these systems might have an important role during infection. Although both systems upregulate pmrA, pmrB and the genes responsible for the addition of aminoarabinose to lipid A[12,13], these systems are differentially activated in response to certain antimicrobial peptides. For instance, the synthetic peptide CP28 activates the CprRS system, whereas the ParRS system seems to be more responsive to the peptide indolicidin[13]. The two systems react similarly to other peptides, such as polymyxin B and colistin, both of which are used to treat P. aeruginosa infections[13]. P. aeruginosa is a useful, but complex, model pathogen for studying the regulation of lipid A modification because structurally divergent forms of lipid A are associated with different types of infection, including lung infections that result in acute bronchiectasis, and chronic colonization of the cystic fibrosis lung[26].

The PhoPQ system has also been shown to mediate modification of lipid A through transcriptional activation of non-coding sRNAs. Many sRNAs modulate the expression of outer-membrane proteins[27-29], but recently the sRNA MgrR of E. coli was shown to regulate lipid A modification[30] (FIG. 2a). MgrR is a transcriptional target of PhoP and is conserved in E. coli and certain Citrobacter, Enterobacter and Klebsiella spp.[30]. This sRNA regulates various genes, including the negative regulation of eptB[30], which encodes an enzyme that transfers phosphoethanolamine to the outer Kdo (3-deoxy-d-manno-octulosonic acid) residue of LPS[7] (FIG. 1c). Although EptB activity results in a modest increase in resistance to the cationic antimicrobial peptide (CAMP) polymyxin B, it is unlikely that this is the main function of this enzyme because conditions that inhibit EptB (through activation of the PhoPQ system) simultaneously induce the other PhoPQ-regulated genes, which confer higher CAMP resistance[30]. In addition to regulating an sRNA, the PhoPQ system itself is controlled by another sRNA, MicA (FIG. 2a), which inhibits translation of PhoP by competitive binding to the ribosome-binding site of the phoP mRNA[31]. MicF represents another example of an sRNA that interacts with lipid A modification enzymes. This sRNA binds to lpxR transcripts, which encode a lipid A deacylase (FIG. 1c; TABLE 1), and increases degradation of the mRNA by exposing regions that are susceptible to RNase E, a major contributor to RNA turnover in many bacteria[32].

Post-translational control

In addition to transcriptional regulation, lipid A modification enzymes are subjected to post-translational control mechanisms. For example, in S. Typhimurium, the PmrAB system orchestrates a delayed negative feedback loop that can be activated by Fe3+ and allows initial uptake of the ion but sets in motion a shift in the cell surface charge that reduces Fe3+ retrieval from the extracellular environment. This feedback loop is established through PmrAB-induced expression of a short peptide, PmrR, which binds to and inhibits the lipid A-modifying enzyme LpxT[33] (FIG. 2b; TABLE 1). LpxT phosphorylates lipid A (FIG. 1c) and thereby increases the overall net negative charge of the outer membrane. However, when LpxT activity is inhibited, other modifying enzymes are capable of transferring amine-containing constituents to lipid A[34], and the outer membrane eventually becomes less negatively charged and so has a reduced affinity for Fe3+ (REF. 33). This mechanism regulates lipid A-modifying enzymes such as LpxT and those encoded by genes that are transcriptionally induced by PmrAB (such as EptA and ArnT). It also prevents the intracellular accumulation of toxic Fe3+ and dampens PmrA-dependent transcription in a delayed manner as a result of the lowered affinity of PmrAB-activating Fe3+ for the membrane. Such small peptide-mediated feedback control of the systems that regulate lipid A modification might be a common theme. In fact, other small peptides, such as SafA (in E. coli) and MgrB (in E. coli, S. Typhimurium and Yersinia pestis), are inner-membrane peptides that interact with the periplasmic domain of PhoQ and activate or repress PhoQ activity, respectively[35,36].

Other post-translational control mechanisms rely on substrate availability. Although the acyltransferase PagP is transcriptionally upregulated by PhoPQ in organisms such as Salmonella spp. and E. coli, its activation is enhanced by environmental stress and intracellular membrane stress[37]. For example, in E. coli, PagP remains dormant in the outer membrane under standard growth conditions. However, treatment with membrane-perturbing agents, such as EDTA, results in displacement of the phospholipid donor substrate of PagP from the inner to the outer leaflet of the outer membrane, placing it in close proximity to the PagP catalytic domain[38]. PagP cleaves the phospholipid substrate, restoring the composition of the outer membrane and increasing the integrity of the permeability barrier by further acylating lipid A[39] (FIG. 2b).

The affinity and activity of lipid A modification enzymes are also modulated by the chemical composition of lipid A. For example, in Yersinia enterocolitica, LpxR deacylates lipid A at 37 °C, but at 21 °C LpxR activity is impeded. At 21 °C, pmrAB and the genes necessary for aminoarabinose addition are induced, and the elevated amounts of aminoarabinose residues on lipid A seem to suppress LpxR activity[40]. In S. Typhimurium, PagL is similarly repressed by aminoarabinose modified lipid A, although this effect seems to be temperature independent[21].

These various forms of post-translational regulation afford quick response times for modification of the outer defence barrier, as the enzymes are already present and primed to function as soon as the appropriate signal is detected. Both transcriptional and post-translational regulation work together, sometimes as functionally redundant mechanisms, allowing Gram-negative bacteria to adapt to diverse environments and thereby ensure their survival. The diversity of regulatory mechanisms also illustrates the importance of the lipid A structure to the membrane barrier.

Host defences that target lipid A

Considering the intimate contact that humans have with bacteria, the frequency of colonization with pathogenic Gram-negative bacteria is astonishingly low. This is largely attributed to the formidable arsenal of host defences that eliminate invading pathogens by recognizing and responding to highly conserved components of infectious agents, known as MAMPs[41]. Because LPS is an essential component of the Gram-negative cell surface, it serves as an effective MAMP to trigger the innate immune system[1]. The host offers a nutrient-rich but perilous environment for a bacterium. For example, intestinal colonization requires a bacterium to journey through the acidic pH of the stomach and encounter toxic compounds such as bile and antimicrobials during transit[42], and the bloodstream is swarming with LPS-binding proteins, antibodies, complement and immune cells primed to detect LPS[1,43-46]. Every point of entry for a bacterium is well defended, but as many of the protective mechanisms rely on lipid A detection, the modification of lipid A affords the bacterium an opportunity to evade the immune system and establish an infection.

Charge-dependent binding of lipid A

Charge-dependent binding of various host molecules (such as CAMPs, platelet factor 4 (PF4) and members of the complement system) to the bacterial cell surface is a major contributor to host protection. CAMPs are amphipathic molecules that are present on mammalian mucosal surfaces, in bodily secretions (such as sweat and saliva) and in phagocytic cells. CAMP production is a highly conserved defence mechanism of most organisms, but the peptides vary widely in terms of their composition and structure. The genes encoding CAMPs are among the most rapidly evolving mammalian genes[47], and it has been hypothesized that this represents an example of co-evolution, wherein the genes are under selective pressure to mutate as a consequence of the evolution of bacterial resistance to the encoded proteins[48]. One of the primary mechanisms proposed for CAMP-mediated bactericidal activity is association of the positively charged peptides with negatively charged lipid A, followed by CAMP insertion into the bacterial membrane and disruption of the membrane potential, leading to cell death[42,48].

PF4 is another positively charged host molecule that is conserved across vertebrates and is released following platelet activation during infection. This molecule binds the bacterial cell surface at the lipid A domain. Antibodies specific for PF4–lipid A complexes are subsequently produced, leading to increased opsonization and phagocytosis of the bacterium[43]. Furthermore, complement proteins also bind lipid A and modulate the activation of the classical complement pathway to promote bacterial clearance[44-46].

Toll-like receptor 4 signalling

Lipid A is recognized by the Toll-like receptor 4 (TLR4)–MD2 (also known as LY96) receptor, one of many pattern recognition receptors (PRRs) of the mammalian innate immune system. This receptor is present on a wide variety of cell types, including monocytes, lymphocytes and endothelial cells[1]. Binding of TLR4–MD2 to lipid A triggers a signalling cascade that leads to inflammation, cytokine production and the eventual clearance of bacteria through recruitment of effector cells, phagocytosis, cytotoxicity and activation of the complement system[49] (FIG. 3). This inflammatory response can be severe, resulting in tissue damage, organ failure and death, especially in cases of sepsis[1]. Unmodified E. coli lipid A, which contains six acyl chains and two phosphate groups, is the strongest known TLR4 ligand, and lipid A modifications can weaken or abolish TLR4 signalling and change the nature of the downstream cytokine profile (see below)[50-53].

Figure 3 |

Toll-like receptor 4–MD2 signalling.

This simplified Toll-like receptor 4 (TLR4)–MD2 signalling schematic illustrates the two responses—the myeloid differentiation primary response protein 88 (MYD88)-dependent and TIR domain-containing adaptor inducing IFNβ (TRIF) (or MYD88-independent) pathways—that can be differentially stimulated on binding of lipid A to the TLR4–MD2 complex. This binding occurs through the association of lipid A with lipid A-binding protein (LBP) and CD14, and leads to the production of cytokines and clearance of the pathogen[49]. The MYD88-dependent pathway leads to the production of pro-inflammatory cytokines, whereas the less inflammatory TRIF pathway occurs after endocytosis of the TLR4–MD2 receptor and stimulates the expression of interferon (IFN)-inducible genes that are important for adjuvanticity but are less inflammatory than those cytokines induced by the MYDD88-dependent pathway.

The detection of lipid A by TLR4 begins with LPS-binding protein (LBP) and CD14 binding to LPS and the subsequent transfer of LPS to the TLR4–MD2 complex. This complex can then signal through two major pathways, which are named according to their adaptor proteins: myeloid differentiation primary response protein 88 (MYD88) and TIR domain-containing adaptor inducing IFNβ (TRIF; also known as TICAM1)[54] (FIG. 3). Severe reactions to LPS are attributed to activation of the MYD88 pathway, which induces the production of pro-inflammatory cytokines such as tumour necrosis factor (TNF), interleukin-6 (IL-6) and IL-12. The less inflammatory TRIF (or MYD88-independent) pathway occurs after endocytosis of the TLR4–MD2 receptor and is characterized by the production of interferon-β (IFNβ) and IFN-inducible proteins such as 10 kDa IFNγ-induced protein (IP10; also known as CXCL10), monocyte chemoattractant protein 1 (MCP1; also known as CCL2), RANTES (also known as CCL5) and granulocyte colony-stimulating factor (G-CSF)[55]. Although it is important for mounting an optimal immune response to pathogens, the TRIF pathway does not lead to severe inflammation. Unmodified LPS from E. coli induces signalling through both pathways, but lipid A modifications can cause preferential recruitment of one adaptor protein over the other[56] (BOX 2). In vivo, bacterial lipid A modifications are also known to affect the potency of TLR4 activation, as described below[7].

Bacterial evasion strategies

Modification of lipid A equips Gram-negative bacteria with an ability to evade immune recognition and survive within a host. First, by changing the overall charge of the bacterial surface through the addition of chemical groups to lipid A, such as the addition of phosphoethanolamine and aminoarabinose, resistance to innate immune effectors (for example, CAMPS and complement factors) increases. Second, changes in the structure of lipid A are important for bacterial pathogenesis because they directly affect recognition by the TLR4–MD2 receptor, and the degree of lipid A acylation and phosphorylation is crucial for LPS recognition by TLR4–MD2 (REF 50). Bacteria can remove acyl chains and phosphate groups to evade detection by this PRR or to shift the type of cytokine response induced[57]. Finally, certain modifications (such as those regulated by PhoPQ) alter the properties of the outer-membrane permeability barrier, which provides resistance to harsh pH and antibiotics, among other stresses.

There are benefits and costs associated with each modification: as bacteria adapt to protect themselves against certain assaults, this might result in the loss of protection against others. For instance, PhoPQ- and PmrAB-induced lipid A modifications increase resistance to CAMPS, but constitutive activation of PmrAB has been shown to reduce the resistance of E. coli to the bile component deoxycholate[58]. Furthermore, PhoPQ-mediated alterations have been shown to lower the resistance of S. Typhimurium to antibiotics in the presence of high Mg2+ concentrations[24]. However, these costs of lipid A modification seem to be outweighed by a number of advantages for bacterial virulence, as exemplified by S. Typhimurium, Helicobacter pylori, Y. pestis, Francisella tularensis and Vibrio cholerae.

Salmonella enterica subsp. enterica serovar Typhimurium

S. Typhimurium is an intracellular pathogen that causes gastroenteritis in humans. As a long-standing model organism for studying bacterial pathogenesis, S. Typhimurium provides prime examples of complex lipid A alterations[7] (FIG. 1d). The bacterium has a diverse lifestyle and fine-tunes the composition of lipid A in response to the surrounding environment[7,59]. During infection, S. Typhimurium penetrates the epithelial lining of the small intestine, invades lymphoid tissue and infects host phagocytes[60]. The unmodified lipid A synthesized by this organism is identical to that produced by E. coli K12 (FIG. 1c). However, survival is promoted in the intestinal lumen and within host cells (where the bacterium encounters CAMPs, low pH and possibly other unknown signals) by activation of the PhoPQ and PmrAB systems, leading to the addition of phospho-ethanolamine and aminoarabinose by EptA and ArnT, respectively, and to acyl chain remodelling by PagP and PagL[19] (FIG. 1d). Commensal and pathogenic E. coli strains also encode these lipid A modification enzymes, but little is known about their regulation in vivo. In S. Typhimurium, another enzyme, LpxO, hydroxylates lipid A (FIG. 1d) as part of a coordinated stress response[61]. The combination of these modifications results in a remodelled outer membrane with a reduced net negative charge and increased integrity, which enhances virulence[9,25].

Helicobacter pylori

The human stomach is the sole niche of H. pylori, and the organism is so well adapted to this environment that it colonizes roughly 50% of the world’s population and can persist for decades within a single host[62]. To survive chronically in the host and remain undetected, H. pylori uses two constitutive lipid A-mediated evasion strategies: repulsion of CAMPs (which are present at high concentrations in the gastric mucosa) and evasion of detection by TLR4 (FIG. 4a). Similarly to most Gram-negative bacteria, H. pylori synthesizes a hexa-acylated lipid A, but displays a tetra-acylated molecule lacking phosphate groups on the bacterial surface. This striking structural difference between the originally synthesized lipid A and the surface-exposed molecule is due to the actions of several enzymes, including dephosphorylation by LpxE and LpxF, addition of phosphoethanolamine by EptA and deacylation by LpxR (TABLE 1). These modifications confer resistance to polymyxin B as well as other biologically relevant CAMPs[63]. The reduced acylation and phosphorylation of lipid A also lead to decreased stimulation of TLR4 and the downstream signalling cascade[50,64,65]. When these modification systems are inactivated through mutation, H. pylori displays hexa-acylated, bis-phosphorylated lipid A (FIG. 4a), which is a strong stimulator of TLR4 (REF. 63). The constitutive lipid A modifications to the acyl chains and phosphate groups are adaptations that allow this bacterium to persist in the harsh gastric environment amidst the several antibacterial components of the innate immune response.

Figure 4 |

Lipid A modification strategies that promote survival in the host.

Helicobacter pylori, Yersinia pestis and Vibrio cholerae have evolved different mechanisms of lipid A modification to aid colonization of their respective niches inside the host. Differences in structure between the pairs of lipid A molecules are highlighted in red. a | The human stomach is the sole niche of H. pylori, and the bacterium has adapted to this environment by constitutively modifying lipid A to a form that resists cationic antimicrobial peptides (CAMPs) and evades the lipid A receptor, Toll-like receptor 4 (TLR4). Surface expressed H. pylori lipid A consists of a tetra-acylated form that lacks the 4′-phosphate group and is substituted at the C1 position with a phosphoethanolamine[63]. Lipid A modification mutants present a hexa-acylated, bis-phosphorylated species that is a TLR4 agonist. b |When residing in the flea vector, Y. pestis produces an endotoxic hexa-acylated lipid A that is a strong immunostimulant in humans. Following transmission to the human host, the bacterium senses a shift in temperature (from 21–27 °C to 37 °C) and synthesizes tetra-acylated lipid A, which escapes detection by TLR4, and immune stimulation is thereby curtailed[70]. c | V. cholerae normally inhabits freshwater, estuarine and oceanic environments and often colonizes marine organisms such as copepods, which are important for cholera transmission. Copepod colonization provides an environment that probably induces lipid A modifications which are crucial for marine and host survival, a prospect that is currently under investigation. The V. cholerae O1 El Tor biotype modifies lipid A with a glycine or diglycine residue, providing resistance to CAMPs, whereas the classical V. cholerae O1 biotype remains susceptible[75].

Yersinia pestis and Francisella tularensis

Y. pestis is notorious for its role in human disease throughout history, causing the Black Death plague that killed approximately one-third of the European population in the fourteenth century[66]. The bacterium has a complex life cycle, colonizing both the flea and human host[67]. This transition between hosts coincides with a switch in the composition of lipid A[68] (FIG. 4b). Inside the flea, Y. pestis grows at a temperature of between 21 °C and 27 °C and synthesizes a hexa-acylated lipid A similar to the highly inflammatory E. coli lipid A that strongly stimulates TLR4 (REF 69). In humans, Y. pestis encounters a temperature of 37 °C and replaces the agonist form of lipid A with a tetra-acylated form that is a TLR4 antagonist. The tetra-acylated form of lipid A is believed to allow the pathogen to proliferate undetected in the bloodstream during the early stages of infection[70,71]. Heterologous expression of the E. coli acyltransferase LpxL in Y. pestis restores hexa-acylated lipid A and strong TLR4-stimulatory properties[70]. Strains that can produce only hexa-acylated lipid A are also avirulent in mice, suggesting that deacylation of lipid A is required for TLR4 evasion and is crucial for Y. pestis pathogenesis[70].

Similarly to Y. pestis, the lipid A of F. tularensis is modified according to temperature, affecting membrane integrity and pathogenesis in the environmental vectors of this pathogen, such as protozoa and arthropods (18–26 °C), and mammalian hosts (37 °C). This bacterium has two homologues of LpxD, an essential lipid A biosynthesis acyl transferase that incorporates longer acyl chains at high temperature (37 °C) than at lower temperatures (25 °C and 18 °C)[72]. A mutant strain that is unable to produce lipid A with longer acyl chains is avirulent in mice and is also more susceptible to antibiotics and CAMPs owing to increased membrane permeability[72]. These two examples of lipid A modifications emphasize the fact that appropriate adaptation in response to temperature is important for altering the fluidity of the outer membrane, and that this alteration is needed to maximize Y. pestis and F. tularensis virulence.

Vibrio cholerae

V. cholerae, the causative agent of the severe diarrhoeal disease cholera, commonly survives on marine crustaceans such as copepods and can be acquired by humans through ingestion of copepods, leading to intestinal disease[73]. The V. cholerae O1 El Tor biotype is currently causing a seventh worldwide pandemic affecting an estimated 3–5 million people, but the previous six pandemics were caused by the classical biotype V. cholerae O1. These two biotypes can be differentiated by their sensitivity to polymyxin B[74], as V. cholerae O1 El Tor is 100-fold more resistant to polymyxin B than the classical V. cholerae O1 strains. This resistance is conferred by an unusual lipid A modification. A three-gene operon in V. cholerae (almEFG) encodes enzymes that transfer a glycine or diglycine residue to a 3′-linked acyl chain of lipid A[75] (FIG. 4c). This is the first described amino acid modification of lipid A, and it has revealed a unique link between Gram-negative and Gram-positive cell wall decoration[75,76], as Gram-positive bacteria can transfer the amino acid alanine to wall teichoic acids and wall lipoteichoic acids and thus reduce the net negative charge of the cell wall[76]. The amino acid modification of lipid A in V. cholerae has little effect on TLR4 activation, suggesting that it has evolved for the distinct purpose of promoting survival in the presence of CAMPs in the marine and human host environment[75].

Host modification of lipid A

Host tolerance of commensal bacteria

The host environment at mucosal surfaces is rich in LPS, as this compound is produced in abundance by commensal Gram-negative bacteria. This means that the immune response at these surfaces needs to be modulated to promote tolerance to commensals but to mount a defence against pathogens. In general, enteric bacteria produce highly stimulatory, hexa-acylated lipid A[77], although decorating the cell surface with such an efficient immunostimulant seems counterintuitive as a strategy for commensal bacteria. In the gastrointestinal lumen, tolerance of commensals is partly attributed to the limited expression of TLR4 on the apical surface of epithelial cells, which means that TLR4 expression in the subepithelial layers is important to restrict infection following damage to the outer layer of cells. On the other hand, Paneth cells in the small intestine have recently been shown to produce multiple antimicrobial proteins in order to control inflammation and infection caused by Gram-negative bacteria, and this response is triggered directly in a TLR4-dependent manner by the presence of bacteria and purified LPS[78]. Paneth cells seem perfectly situated to respond appropriately to commensal bacteria, producing an antimicrobial moat that keeps bacterial numbers balanced to avoid excessive inflammation but permit enough commensal colonization to impede invasion by pathogens[78].

Host detoxification of lipid A

When the innate immune response has been triggered by Gram-negative bacteria that have successfully breached the mucosal and epithelial barriers, LPS must be inactivated or removed from the bloodstream to reset the immune response for subsequent LPS exposure. Various host response mechanisms are involved in clearance of and tolerance to LPS from commensal and pathogenic bacteria. Covalent modification of lipid A by host enzymes is one mechanism to inactivate lipid A. The mammalian lipase acyloxyacyl hydrolase (AOAH) is produced by macrophages, dendritic cells and neutrophils and cleaves all secondary linked acyl chains of lipid A (that is, the O-linked acyl chains that are attached to one of the four primary acyl chains linked directly to the sugar backbone)[79]. This renders the LPS molecule less inflammatory owing to impaired interactions with TLR4 (REFS 50,64). Exposure of AOAH-deficient mice to LPS results in prolonged tolerance or unresponsiveness to LPS; this effect can last for months, thus limiting the capacity of the host to react to a secondary exposure. By contrast, wild-type mice are able to recover and respond normally to LPS within 5–10 days[80]. AOAH activity increases during LPS challenge in rabbits and mice, suggesting that this enzyme has a primary role in LPS detoxification[79,80].

Host alkaline phosphatases have also been shown to modify LPS through the removal of phosphate groups. Whether these phosphatases act on both lipid A and the sugar chains of LPS requires further study, although several studies have shown that alkaline phosphatases interact with LPS in vivo[81-83]. Similarly to deacylation, removal of the lipid A phosphate groups results in reduced TLR4 stimulation and reduced LPS potency[50]. Recently, a direct link between LPS and one of these phosphatases (intestinal alkaline phosphatase (Iap; also known as Alpi1)) was identified in zebrafish. This particular phosphatase probably cleaves phosphate groups from the lipid A domain of LPS, and it controls the inflammatory response to gut microorganisms. Zebrafish with mutations in iap are sensitive to LPS, whereas in wild-type zebrafish, LPS induces TLR4-dependent signalling that upregulates the transcription of iap, effectively lowering LPS toxicity[83].

Lipopolysaccharide scavenging

Many other host molecules have been identified that associate with LPS (usually at the lipid A domain) and assist in reducing endotoxin concentration in the body and trafficking LPS for disposal. These molecules have been shown to reduce LPS toxicity by direct binding to LPS or by competing with it for association with essential cofactors of TLR4 (CD14 and LBP)[84]. For example, pre-incubation of the antimicrobial peptide LL-37 with LPS results in reduced cytokine induction in exposed THP-1 cells compared with exposure to LPS alone, suggesting that LL-37 is an LPS scavenger[85]. The abundant plasma protein β2 glycoprotein I (β2GPI) forms complexes with LPS through an LPS-binding domain that is well conserved across vertebrates[86]. Binding of β2GPI to LPS causes a conformational change in β2GPI that leads to the association of these complexes with monocytes and subsequent clearance of LPS[87]. LBP is important for binding LPS and recruiting it to CD14 and TLR4 to induce immune signalling, but it has also been implicated in a mechanism of neutralizing LPS by transferring the endotoxin to high- and low-density lipoprotein and to chylomicrons, enhancing clearance of LPS following uptake by the liver[88,89]. The direct binding of host factors to lipid A, as well as the effects of lipid A modifications on the above mechanisms, have not yet been investigated, but further studies could reveal additional evasion strategies that are used by Gram-negative bacteria to promote pathogenesis.

Effects on other cellular processes

Recent studies have revealed links between lipid A modifications and diverse processes within the bacterial cell. These associations support the notion that lipid A modifications are crucial for bacterial pathogenesis and survival.

Multitarget lipid A modification enzymes

In Campylobacter jejuni, the enzyme EptC modifies both lipid A and a structural protein required for the assembly of flagella. EptC is a homologue of E. coli EptA, and both enzymes transfer a phosphoethanolamine residue to the 1′ and 4′ positions of lipid A (TABLE 1), masking both phosphate groups and promoting CAMP resistance. Interestingly, in C. jejuni, EptC also transfers an essential phosphoethanolamine to FlgG (a flagellar rod protein[90]), and EptC mutants lack wild-type motility and produce fewer flagella, which are required for virulence. This promiscuous C. jejuni enzyme also modifies carbohydrates of the LOS core and glycosylated proteins[91].

Recycling of biosynthetic intermediates

During the assembly of bacterial polymers (for example, peptidoglycan, the capsule and LPS), a wide array of glycoconjugates must be trafficked across the inner membrane[92]. These glycan intermediates are assembled on a universal carrier lipid known as undecaprenyl phosphate (C55-P)[93]. The C55-P transports the glycan precursors via a pyrophosphate linkage (C55-PP–precursor), and when the precursor is removed, the carrier lipid is released as a pyrophosphate and requires dephosphorylation in order to be recycled[93]. Within the periplasm, LpxT transfers a secondary phosphate group onto lipid A at the 1 position (FIG. 1c) and uses C55-PP as a donor[94]. This finding represented a novel and unexpected link between the modification of lipid A and the recycling of C55-PP. Furthermore, the connection between these systems implies that C55-PP could serve as a high-energy phosphate donor for various periplasmic components, although these potential targets are largely unidentified.

Activation of outer-membrane proteins

As a membrane component, LPS is more than just a platform for activities at the cell surface[95]. It also has a role in the activation of omptins, a class of proteases that is widespread among Gram-negative enteric bacteria. The activity of one member of this class, OmpT, has been evaluated in the presence of hexa-acylated and tri-acylated lipid A forms, and the protease was found to be active only in the presence of hexa-acylated lipid A[96]. Interestingly, another omptin named plasminogen activator (Pla), which is central to Y. pestis pathogenesis, is more active in Y. pestis grown at 37 °C and producing tetra-acylated lipid A than in cells grown at 20 °C and producing hexa-acylated lipid A[97].

Outer-membrane vesicles and toxin delivery

Lipid A modifications also affect the delivery of proteins and toxins that are associated with the outer membrane and outer-membrane vesicles (OMVs). In organisms such as Porphyromonas gingivalis and P. aeruginosa, negatively charged LPS is enriched in OMVs[98,99]. Furthermore, in P. gingivalis, the lipid A found in OMVs is more deacylated than that present in whole-cell membranes, suggesting that lipid A modification facilitates vesicle formation or the sorting of outer-membrane proteins that are packaged into vesicles[98]. In P. aeruginosa, OMV formation is stimulated by a small signalling molecule called Pseudomonas quinolone signal (PQS), which is packaged into vesicles and induces vesicle formation in P. aeruginosa as well as other bacteria[100]. The exact mechanism of OMV formation is not understood; however, PQS was found to interact with the 4′-phosphate group and acyl chains of lipid A, indicating that modification of lipid A could influence the efficiency of vesicle shedding[101].

The structure of lipid A and the Kdo residues of LPS also affect the delivery of toxins. For example, two similar toxins — cholera toxin (CT), produced by V. cholerae, and heat-labile enterotoxin (LT), produced by enterotoxigenic E. coli (ETEC) — are both secreted, are structurally similar and bind the same host cell receptor. However, CT causes a more severe diarrhoeal disease than LT[102]. Although the topic is controversial[103], the difference in toxin severity seems to be due to the partial sequestration of LT, as it binds to Kdo-lipid A on the bacterial cell surface after secretion. By contrast, the Kdo sugar attached to lipid A in V. cholerae is phosphorylated by Kdo kinase (KdkA) (TABLE 1), and this modification inhibits CT association with the outer membrane, thereby allowing robust toxin secretion[104].

Conclusions and future prospects

The wide variety of lipid A modifications equips Gramnegative bacteria for survival in their respective niches in the host and other environments. Other functions of lipid A are highlighted by the interplay of lipid A modifications with a diverse set of cellular processes, and both of these facets of lipid A biology emphasize the importance of a continued effort to understand the regulation, benefits and host response to lipid A in its numerous modified forms.

However, many questions remain unanswered. The continued development of animal models is paramount to more closely mimic the host infection site. Considerable progress has been made in this regard by the recent generation of a transgenic mouse expressing the human TLR4 receptor[105]. Furthermore, there is much to learn about how lipid A and its modifications affect the outcome of polymicrobial infections, with recent studies suggesting that LPS has an important role within such communities. For instance, purified LPS from intestinal commensal bacteria has been found to promote replication and transmission of viruses[106,107]. Pathogenesis of orally acquired poliovirus, which replicates in the intestine before spreading to cause a severe systemic infection, is supported by the intestinal microbiota and purified LPS[107]. Mouse mammary tumour virus (MMTV) manipulates the innate immune response by binding LPS and eliciting TLR4-dependent production of IL-10, a cytokine that is required for MMTV persistence[106]. Although these viruses bind LPS and this enhances infectivity in vitro, how lipid A modifications affect these phenomena is currently unknown.

Improved methods for studying lipid A modifications in vivo could also help elucidate any potential links between modified lipid A and diseases involving TLR4, such as inflammatory bowel disease, diabetes, rheumatoid arthritis and atherosclerosis[108]. Important insights into human TLR4 polymorphisms might now be possible using the humanized mouse model[105], and it should also be possible to examine how different variants of TLR4 affect susceptibility to infection by organisms with modified lipid A. This could potentially lead to more personalized medical treatment through the development of a new range of modified TLR4 agonists that could serve as customized immunomodulatory agents to induce or repress particular cytokine responses.

It remains a mystery why lipid A is essential in virtually all Gram-negative organisms. Furthermore, why lipid A biosynthesis is well conserved only to have complex (and varying) modifications occur later, sometimes constitutively, is not understood. How the modification enzymes are distributed and organized within the membrane is also unclear. For example, it is not known whether lipid rafts exist in bacteria and whether some or all of the modification enzymes are enriched at these sites. Lipid A molecules with certain modifications could also be specifically localized in the membrane. Such an organization could have a major influence on the activation of key outer-membrane proteins or on OMV formation and protein sorting.

With further study, inhibitors of modification enzymes could be developed for use as antibiotics or in combination with currently used antibiotics. Although immense progress has been made in our understanding of lipid A modification systems in Gram-negative bacteria and their importance for pathogenesis, there is much work yet to be done.